Analysis of the sports action recognition model based on the LSTM recurrent neural network
With the rapid growth of motion data, the traditional motion recognition algorithm is faced with the problem of insufficient processing ability. To solve this problem, a method based on gradient descent optimization (GDO)–long short-term memory (LSTM) is proposed to meet the needs of sports action r...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
De Gruyter
2025-02-01
|
| Series: | Nonlinear Engineering |
| Subjects: | |
| Online Access: | https://doi.org/10.1515/nleng-2024-0050 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | With the rapid growth of motion data, the traditional motion recognition algorithm is faced with the problem of insufficient processing ability. To solve this problem, a method based on gradient descent optimization (GDO)–long short-term memory (LSTM) is proposed to meet the needs of sports action recognition. Based on the experiment of sports data set of students in Hainan University, the experiments of skipping rope, swimming, skating, and shotput were carried out extensively. The total number of experiments were 77, 94, 72, and 85. The experimental results show that the accuracies of GDO-LSTM in sports action recognition were 98.7, 100, 100, and 94.1%, respectively, which was superior to that of the three-axis gyroscope (80.5, 40.4, 23.6, and 100%). These results show that the algorithm can effectively improve the accuracy of sports action recognition and has wide application potential. |
|---|---|
| ISSN: | 2192-8029 |